Cognitive scientists have yet to create a functional, working model of the human brain. Our computers are way too slow, and our understanding of the brain is way too primitive. But that doesn't mean we can't create a model of the functioning brain at a much smaller scale. And in fact, this is exactly what a group of researchers from Waterloo University in Canada have done — what they're calling the largest functioning model to date.
It's called Spaun, short for Semantic Pointer Architecture Unified Network — a system that is powered by over 2.5 million simulated neurons. And fascinatingly, every neuron is an entity unto itself; each of them captures the biological detail of a real, biological neuron — including which neurotransmitters are used, how voltages are generated in the cell, and how they communicate.
Sub-networks then mimic the structure and functions of several anatomical areas, including perception, action, and cognitive control. In other words, it has a virtual prefrontal cortex, basal ganglia, and thalamus — what enables Spaun to ‘think' about its environment and respond to the patterns it encounters. As a result, it it can simulate complex behaviors such as thinking, remembering, seeing, and interacting with its environment.
Researcher Chris Eliasmith and his colleagues are using the simulated brain to perform a series of simple, yet distinguable tasks — what's comparable to a human's ability to shift from task to task. Specifically, Spaun has been given problems involving perceptual, cognitive, and motor tasks. It uses its network of neurons to answer questions given to it via a small screen, and it draws its answers with a mechanical arm. So the system is not just cerebral; the researchers have endowed it with a kind of physicality.
Spaun is capable of completing any one of eight different tasks, each of which requires a different set of cognitive skills. So, for example, it can memorize a list of numbers, or recognize an object.
And Spaun likes to take its time; it takes about 2.5 hours of computer time for every second of simulation — a consequence of the system's massively reduced scale relative to the human brain.
According to the researchers, Spaun could eventually be used to understand how changes to the brain affect changes to behavior. For example, a gradual loss of neurons can be modeled against cognitive decline in humans — what is a similar deficiency.
In addition, the researchers hope to use Spaun as a way to improve machine intelligence as a whole, such as controlling the flow of information through a large system that's attempting to solve a difficult task.
It's important to note that Spaun is driven exclusively by the interplay of algorithms and its simulated neurons, and not by any kind of conscious awareness. Its ability to contextualize and solve problems is not an indication of any kind of internal reflection about what it's supposed to do or how it relates to its environment.
That said, it's quite possible that a real, conscious mind taps into these problem solving processes as it engages in cognition. So while it's not a complete model by any extent of the imagination, it's very likely an important partial model.
But until neuroscienctists develop a proper model for consciousness, systems like Spaun are merely script-driven automatons — even if it is a system comprised of 2.5 million artificial neurons.
You can read the entire study at Science.